This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: Hypothesis testing Terminology
• Null hypothesis – e.g. H0: µ = 1 0 , H0: p = .5 , H0: µ1 = µ2 • Alternative hypothesis – e.g. H1 : µ ≠ 1 0 , H1: p < .5 , H1 : µ1 > µ2 •
•
•
•
•
• Type I error?
Type II error?
Significance Level?
Critical Region?
pvalue?
Power?
2 Type I and II errors
• Type I error
– Reject H0 when H0 is true • Type II error
– Accept H0 when H0 is false • Probabilities of Type I and II errors should
be as small as possible
– Fixed sample size
• reducing probability of type I error >increases
probability of type II error
3 Hypothesis testing approach – significance
level
• FIX the maximum allowable probability
of Type I error = α, significance level of
the test
• Test is then arranged to minimize Type
II error
• Most expensive error should represent
Type I error
• Type I error can be fixed
• Under more control than type II error
4 Hypothesis testing example
• Large sample size σ
x ~ N (µ , )
n
2 • One sided test
– H0:µ = µ0 vs H1:µ > µ0 – e.g. H0:µ = 1 0 vs H1:µ > 1 0 Ｉ Ｉ Ｉ Ｉ Ｉ Ｉ is true σ2
n α µ0
Critical region: Reject
H0 µ0 + k x Hypothesis testing approach – significance
level
• FIX the maximum allowable probability
of Type I error = α, significance level of
the test
• Reject H0 if test statistic lies in the
critical region
x − µ0
> zα
1
σ
n
7 Computer Output (pvalue)
• pvalue is a measure of exactly where the
teststatistic lies in the critical region
– Smaller pvalue > more significant result • Reject H0 if pvalue is less than α pvalue measures
how significant a
result is pvalue µ0 x Hypothesis testing approach – power
• Standard tests are designed to
minimize
probability of Type II error = β • Power of the test = 1 – β
• Standard test designed to have
maximum possible power (depending
on effect size)
• Recommended power ∼ 0.8
• Fixed n  increase power by relaxing α?
•
10 ...
View
Full
Document
 Spring '09
 ValerieOzaki
 Null hypothesis, Hypothesis testing, Statistical hypothesis testing, Statistical significance, Type I and type II errors, Statistical power

Click to edit the document details